Looking into this.

Thank you!

Thank you very much for the report!

Could you please provide full example, so that other people can easily copy and reproduce the situation?

If you want, we can add this example to the VPA vs. MCT comparison here:

Symbolic Math Toolbox (VPA) vs. Multiprecision Computing Toolbox

We will also include backlink to your page, of course.

Thank you for the suggestion. 

At the moment, we have no plans to implement this functionality. Because it is out of scope of standard MATLAB and its usage scenarios. The link you provided has not been updated for many years (with zero downloads).

We will reconsider this, if this request will be supported by other users.

Jon, thank you for the suggestion.

There are a lot of other general tasks we need to do regarding better web-infrastructure, license management, automatic updates, etc. etc. Hard to select which one is the most important. 

Now we provide previous versions by request. This gets the job done. Besides this gives us a chance to communicate with users, to hear feedback on toolbox, etc. Running cost is also important for non-profit like us.

Hi Jon,

Please let me know what versions you are looking for (by email).

I will send you the download links.

We cannot keep previous versions on our server permanently due to size limitations (toolbox bundle is ~100MB for each platform).

This is the same idea as 'precomputeLU', but we introduced it years before 'decomposition'.

We will support 'decomposition' in future versions, of course.

Dense matrices is way more easy/faster to handle.

[L,U,P] = lu(A);  % pre-compute dense LU beforehand

...
x = U\(L\(P*b));  % re-use it in a loop for different right-hand sides
...

In future versions we will support same syntax for sparse matrices (and also will keep the precomputeLU)

This is definitely possible with usual MATLAB functions, e.g.:

function F = mpfun(x)
        x = mp(x)      % convert input to 'mp'
        ...            % compute F using 'mp'
        F = double(F)  % return 'double'
end

S-function is complied MEX module and it cannot use "mp" objects directly in C/C++ code.

But you can put 'mp'-related code to separate MATLAB function (like above) and call it from within MEX/S-function using mexCallMATLABWithTrap or mexCallMATLAB.

Sorry, we didn't include it so far. 

Just remove toolbox directory from search path in MATLAB.

This aspect is controlled by user - toolbox doesn't change search path automatically.

You included toolbox to MATLAB search path permanently after toolbox installation - so this is not a bug of toolbox.